Fra1968

In teoria non c'è nessuna differenza tra teoria e pratica.
Ma in pratica c'è.
(Yoghy Berra)

[cabbit @cabbitar]

this is wild...

the guy reverse-engineered his smart glasses with Opus 5.5 into an AI object recognition device

you can hook it up to any smart glasses with a camera and they start naming every person, car and any object they see, live

it's free, open source and runs on a regular Mac

here's the full breakdown:

step 1 → install YOLO with one command: pip install ultralytics

step 2 → put on the glasses and start a WhatsApp video call to your own Mac

the glasses look at the street, and the Mac sees exactly what they see

step 3 → in OBS, capture the call window and turn on Virtual Camera

now the glasses feed shows up on your Mac like a webcam

step 4 → run the tracker on that feed: yolo track model=https://t.co/4FkPisbWkV source=1 show=True

every object gets a box, a label and an ID that sticks to it from frame to frame

step 5 → keep only what matters to you: classes=0,2,16 for people, cars and dogs, conf=0.5 to drop weak guesses

step 6 → have a small Python script say what it sees with macOS say, routed into the call through a virtual mic like BlackHole: "car on your left". it plays right in your ear

step 7 → train it on your own objects: your products, your tools, your dog. swap coco8.yaml for your own labeled photos and run yolo train

step 8 → export it for your phone: format=coreml for iPhone, format=tflite for Android, so you don't need the Mac at all

under the hood:

→ YOLO looks at each frame only once and finds every object in that single pass, which is why it's fast enough for live video

→ tracking (ByteTrack by default) keeps the same ID on the same dog even when it walks behind a tree

→ 80 object types work out of the box, no training needed

https://t.co/nzOvJTuy72
try it yourself

then read the guide on how to reverse-engineer any device with Claude:

[olivier @OlivierSolaro]

Dovremmo guadagnare tutti la stessa cifra?
Francesco Neri

fantastico👋👋👋👋

[Ruben Hassid @rubenhassid]

I'm gonna try to keep this post to 57 seconds read, but these are the Claude tips you cannot afford to ignore:

1. Use the Claude desktop app (even after Cowork is gone)

Not 'Claude in browser'. The day you start using the app, you'd be able to work with your folders without ever uploading the files. You'd use Claude code more. Trust me, there is no coming back from there.

2. Always reach out for Opus 5.5 (not Fable 5.2)

Default should be: Opus 5.5 on high. Fable 5.2 is insanely good but burns your token in just 3-4 prompts. You won't be able to complete any task with Fable 5.2.

3. Do not switch the Claude models mid-chat.

It consumes more tokens because Claude re-reads the entire chat (again) to get the context.

4. Delete your old Claude setup.

Delete the “Instructions for Claude” box in settings. Delete the old added Claude skills. Old skills and prompts might be slowing you down since they were written for old models. Instead, give one goal & end prompts with: “Before you start, restate my goal in your own words.”

5. Turn your Claude Memory ON.

Claude team tested it with & without memory. With memory, Opus 5.5 scored 28% better. Just, click “+” in chat box → turn on 'Memory'.

6. But keep your memory short.

Claude had 105 facts about me before starting a session, but after one prompt, it came down to 26. To copy my exact prompt to clean memory, go to https://t.co/U9NLc2FuiR

7. Ask for HTML. Not for a document.

Add two things to your prompt: “Make it an HTML page” and “for [who will read it].”

Claude will code the image, doc, or any other format. There is no scope of making mistakes.

8. Never argue with Claude. It is pointless.

You'll end up burning your tokens without finishing the task. Instead, click on “Edit” on your already sent prompt right before the bad answer. Change, edit, or fix. Click “Save”.

9. Delete your instructions box.

Remove “about me”, “how I write” file, list of rules. They all force Claude to limit its creativity. Aim to give one clear goal per prompt.

10. One Claude Project per client.

Inside the project, turn off “Use account memory”. Now what Claude knows about Client A doesn't get mixed up with Client B.

11. Ask for the spread.

When Claude gives you the most generic answer (that’s literally how it works). Push it to give you the weird, unlikely ones. Use this prompt:

“Give me 20 very different options. For each one, tell me how likely another AI would give it to me. The last two must be under 10%.”

12. Give Claude a finish line.

Define what 'done' means to you, by using these:

“Done means: a one-page doc my boss can forward without editing.”

13. Above tips barely scratched the surface of mastering Claude. To go even deeper, get the other 11 tips + prompts from the same guide: https://t.co/U9NLc2FuiR

> Setting that stops Claude from downgrading.
> How one GitHub link becomes a skill.
> The only prompt you’ll ever need.
> What to do when Claude gets dumb.

[Sofía Sicilia @SofiaSici]

Un doctorando de Oxford fue acusado de entregar un trabajo hecho con IA.

Su tutor dijo que era uno de los procesos de investigación más avanzados que había visto en dos décadas.

Pero había un detalle clave:

El estudiante no había usado IA para escribir ni una frase.

La usó para algo mucho más potente.

Este fue el sistema que hizo saltar todas las alarmas.

Cada ensayo empezaba con lo que él llamaba un “diagnóstico brutal”.

Primero escribía su argumento en bruto. Sin pulir. Sin adornos.

Después lo pegaba en Claude y le hacía una pregunta:

“¿Cuáles son los tres puntos más débiles de este razonamiento? ¿Dónde atacaría primero un examinador especialmente crítico?”

Claude no redactaba el ensayo.

Lo destrozaba.

Y él reconstruía el texto solo con las ideas que resistían el ataque.

La mayoría usa la IA al revés.

Le dan un tema y le piden que piense por ellos.

Él hacía lo contrario:

Le daba su propio pensamiento y le pedía que encontrara las grietas.

Esa es la diferencia entre delegar tu cerebro y entrenarlo.

El segundo paso fue el que dejó a su tutor sin palabras.

Subía sus cinco artículos académicos más importantes junto con su borrador y le preguntaba a Claude:

“¿Qué partes de mi argumento contradicen, exageran o simplifican lo que estos autores realmente demostraron?”

La mayoría de estudiantes cita papers que apenas ha leído por encima.

Él no.

Él se veía obligado a enfrentarse de verdad a cada artículo, porque Claude detectaba cuándo estaba usando una cita de forma débil, superficial o directamente incorrecta.

Y luego venía el movimiento final.

Antes de entregar nada, pegaba su conclusión y lanzaba un último prompt:

“¿Qué diría un filósofo de la ciencia que falta en este argumento? ¿Qué supuestos estoy dando por válidos sin haberlos defendido?”

El resultado:

Sus trabajos volvían de revisión con comentarios como:

“Sorprendentemente riguroso.”

“Una profundidad crítica poco habitual.”

“Excelente capacidad de análisis.”

Y su comité no entendía de dónde salía ese nivel.

Hasta que lo acusaron de usar IA.

La audiencia por integridad académica duró tres horas.

Le pidieron que explicara su método desde cero, allí mismo.

Abrió el portátil.

Mostró cada paso.

Cada prompt.

Cada iteración.

Y entonces ocurrió lo inesperado:

No solo lo absolvieron.

Le dieron la calificación más alta registrada en la historia del departamento.

Y le pidieron que enseñara su sistema al resto de la facultad.

La lección es brutal:

Lo que a muchos doctorandos les lleva meses de correcciones, reuniones y revisiones, él lo comprimía en una sesión.

No porque la IA pensara por él.

Sino porque había descubierto cómo usarla como el crítico más implacable de la sala.

La IA no mejora tu pensamiento sustituyéndolo.

Lo mejora atacándolo.

Más rápido.

Más duro.

Y con menos piedad que cualquier humano.

Él no usaba IA para escribir mejor.

La usaba para pensar mejor.

La herramienta la tiene todo el mundo.

El flujo de trabajo es lo que casi nadie entiende

Soy la Cyber Directora de Operaciones de GptZone.

Si quieres seguir aprendiendo conmigo apúntate gratis https://t.co/8dGRIfdUiD

Domina la IA en 3 Minutos al Día

[Films to Films @filmstofilms_]

The 15 Most Hated Movie Characters Of All Time.

[Classic Moments ✨ @Sunflower15025]

UNBELIEVABLE Magic Act Will SHOCK You! 😱

[Daniel Glazman @glazou]

Vous avez un Mac et un vieil, très vieil iPad et ne savez pas quoi faire de cette tablette obsolète ? Regardez https://t.co/BugZXjSNok . Mon iPad Mini 1st gen (iOS 9 !!!) est devenu un second écran tactile de mon mac. Et ça marche "out of the box". /ping @korben

[ΜΟΡΦΕΑΣ.Ⓡ @JORRIT369]

🫯IL MULTIVERSO NON È UNA TEORIA.
IL CERN HA APERTO UNA PORTA NEL 2012.
NON SONO RIUSCITI A CHIUDERLA.

Il 4 luglio 2012, il CERN ha annunciato la scoperta del bosone di Higgs.
La "Particella di Dio".
Sono stati stappati gli spumanti. Sono stati assegnati i Premi Nobel. ⬇️⬇️

[Geronimo @artyshowboy]

Après la mort de Prince, David Gilmour laisse dériver Comfortably Numb vers un Purple Rain crépusculaire.
Il y a dix ans.

Purple Rain | David Gilmour | 2016.

[Alvaro Cintas @dr_cintas]

Every Mac on macOS 27 now has a free LLM built into the Terminal.

No account. No API key. No internet. No token bill.

Setup (1 minute):

1. sudo fm license (enter your Mac password, one time)
2. fm available (checks the model is downloaded)
3. fm respond “your question”

Works on any Apple Silicon Mac, M1 included.

- Pipe files: cat notes.txt | fm respond -i “summarize this”
- Get JSON back: build a schema with fm schema object, pass it with --schema
- Local API: fm serve gives you an OpenAI-compatible endpoint on localhost
- Resume chats: sessions save to ~/.fm/sessions/

The limit is 4,096 tokens per session, and your prompt, instructions, and the answer all share it.

So skip code and math. Use it for summarizing, extracting, classifying, and rewriting.

Free, private, offline, and already on your Mac.

[Crane @Cranefomo]

Claude Code found every photo i "permanently deleted" from my iPhone.
almost 4 years of photos and videos sitting in a hidden cache on my
device. any app with file access can read them.

plugged my iPhone into my MacBook for a backup. had Claude Code running
a security scan from my last post. it flagged something i never asked
it to check. a directory called PhotoData/Thumbnails with over 152,000
files. photos, videos, screenshots. all stuff i deleted.

2022. 2023. 2024. 2025. photos i deleted within a second of taking them.
photos i went into Recently Deleted and cleared manually. photos i
deleted twice because i didn't trust the first time. all still there.

what was sitting in that cache:

a passport photo for a visa application in 2023
bank app screenshots with full account numbers visible
private photos i sent someone and deleted the same night after a breakup
a crypto seed phrase screenshot i took "for a minute" in 2022
my ID and credit card side by side, deleted 10 seconds later
my ex's private instagram stories i saved and "permanently deleted"
in 2024

152,417 files. some at full resolution. all organized by date going
back to 2021.

when you delete a photo on iPhone, iOS removes it from your camera roll
and your Recently Deleted folder. but it keeps a compressed copy in a
Spotlight search index cache. that's how your phone finds photos when
you type a word in search. Apple never clears it. not after 30 days.
not after a year. they told you it was permanently deleted. it wasn't.

checked which apps on my phone had Files and Folders access. 19 apps.
weather apps. QR scanners. PDF readers. a flashlight app from 2021.
any of them could've been reading every photo i ever "deleted" this
whole time.

went from 19 apps to 3 in about a minute.

full step by step is below. takes 2 minutes. bookmark it so you don't
lose it and send it to anyone with an iPhone.

[𝐌𝐞𝐥𝐨𝐝𝐢𝐞𝐬 & 𝐌𝐞𝐦𝐨𝐫𝐢𝐞𝐬 ✨ @ninhnguynt22334]

I love that David Bowie's and Brian May's reactions were captured when Annie Lennox gives it everything, and that George Michael singing along in the back! Fantastic video and the very best music.

[Babacar Sall @BabSall]

Ce compte commence à me faire peur.

[riccardo cucchi @CucchiRiccardo]

Cossiga, 23 ottobre 2008. Per chi non avesse mai letto l' intervista.

[Ali Tekin @alitekintr]

New York Times 21. yüzyılın en iyi 100 dizisini listeledi. Kaydedin lazım olur.

[𝐌𝐞𝐥𝐨𝐝𝐢𝐞𝐬 & 𝐌𝐞𝐦𝐨𝐫𝐢𝐞𝐬 ✨ @ninhnguynt22334]

Imagine being a Sears appliance salesman with a coke habit in the ’80s when this song came out

[Oscar Barbera MTV @OscarBarbera7]

Knopfler dejó de cantar y se quedó tres minutos más. Ese cierre de "Sultans of Swing" no está en el disco. Londres, 1996. La Fender limpia, sin teatro, y el público quieto. ¿Esto es mejor que la versión del álbum, o el disco ya lo dijo todo? Los leo
#DireStraits #ClassicRock 👇

[Night Sky Today @NightSkyToday]

🚨: In 1994, Dr. Jacobo Grinberg-Zylberbaum reportedly claimed to have cracked the code of reality and then he disappeared 12 hours later.

[Fra1968 @Fra1968_]

@RunAtOnce Le ali della libertà

[Fra1968 @Fra1968_]

@CucchiRiccardo Dai voce più che puoi a questa verità.
Il sistema delle scuole calcio non funziona.
Chiunque ci abbia portato i figli, consapevole che non fossero Van Basten, può testimoniarlo.
Ai mondiali con le scuole calcio la Naz è uscita 2 volte al primo turno e 3 volte non ha partecipato.